Folding of Dimeric Methionine Adenosyltransferase III

Methionine adenosyl transferase (MAT) is an essential enzyme that synthesizes AdoMet. The liver-specific MAT isoform, MAT III, is a homodimer of a 43.7-kDa subunit that organizes in three nonsequential α-β domains. Although MAT III structure has been recently resolved, little is known about its folding mechanism. Equilibrium unfolding and refolding of MAT III, and the monomeric mutant R265H, have been monitored using different physical parameters. Tryptophanyl fluorescence showed a three-state folding mechanism. The first unfolding step was a folding/association process as indicated by its dependence on protein concentration. The monomeric folding intermediate produced was the predominant species between 1.5 and 3m urea. It had a relatively compact conformation with tryptophan residues and hydrophobic surfaces occluded from the solvent, although its N-terminal region may be very unstructured. The second unfolding step monitored the denaturation of the intermediate. Refolding of the intermediate showed first order kinetics, indicating the presence of a kinetic intermediate within the folding/association transition. Its presence was confirmed by measuring the 1,8-anilinonaphtalene-8-sulfonic acid binding in the presence of tripolyphosphate. We propose that the folding rate-limiting step is the formation of an intermediate, probably a structured monomer with exposed hydrophobic surfaces, that rapidly associates to form dimeric MAT III.

The classical investigation of Anfinsen and co-workers (1) demonstrates that the sequence of amino acids that comprise a polypeptide encodes the information required to reach the correct three-dimensional fold of a protein. However, the mechanism by which the amino acid sequence directs the folding of a protein is still unknown. Typically, protein folding studies have concentrated on small, single domain proteins that fold rapidly and avoid aggregation (2)(3)(4). However, in living cells, most proteins have more than 100 residues with multiple folding domains and subunits. These large proteins are not likely to follow all of the folding principles established for smaller proteins. They probably have more complex folding rules so that models based on studies of small globular proteins may not fully explain their folding and assembly (5). Folding of multidomain or multisubunit proteins usually involves the formation of partially folded intermediates. Although the basic principles governing protein folding are probably general, folding of oligomeric proteins may be a two-state process or involve monomeric or multimeric intermediates (6 -9). Additionally, association may occur from completely folded monomers (10 -13), or alternatively, final folding steps occur after association of partially folded subunits (14). All these different mechanisms make it difficult to derive a global mechanism to describe the folding of oligomeric proteins, and consequently each protein must be studied individually. Defining the complete folding pathway requires the identification and characterization of all stable and transient conformational states that exist along the folding and unfolding pathways.
S-Adenosylmethionine (AdoMet) 1 is an essential metabolite in all cells. AdoMet is the most important biological methyl group donor and is a precursor in the synthesis of polyamines (15)(16)(17). Additionally, it has been recently demonstrated that reduction of AdoMet levels in liver is associated with liver injury and hepatocarcinoma (18). Methionine adenosyltransferase (MAT) (EC 2.5.1.6) catalyzes the only known AdoMet biosynthetic reaction (19,20). In all organisms studied so far, from Escherichia coli to humans, sequence homologies show that MAT is an exceptionally well conserved enzyme through evolution (17). In mammalian tissues three different forms of MAT (MAT I, MAT III, and MAT II) have been identified that are the product of two different genes (MAT1A and MAT2A, respectively) (21)(22)(23)(24)(25). The gene MAT2A encodes a 396-amino acid catalytic subunit (␣ 2 ) expressed in all mammalian tissues (16,26). MAT II consist of ␣ 2 catalytic and ␤ regulatory subunits, although the stoichiometry of the oligomer has not been yet established. The gene MAT1A is expressed mainly in adult liver and encodes a 395-amino acid catalytic subunit (␣ 1 ) that organizes in dimers, MAT III, and tetramers, MAT I (17,20). The reason for the presence of these two different isoenzymes in liver has not been yet elucidated, although it may be an adaptation to the metabolic requirements of the liver. The liver has the highest specific activity of MAT, which agrees with the observation that up to 85% of all methylation reactions and as much as 50% of methionine catabolism occur in this tissue ( I and  MAT III isoforms, MAT III has been considered the liverspecific enzyme. While MAT I, similar to MAT II in extrahepatic tissues, may maintain the basal AdoMet levels required  by cells, MAT III would be responsible for the clearance of methionine after a load of this amino acid. E. coli and liver MAT have been crystallized and their structure solved, showing that the tetramer structure of both enzymes is essentially the same (29 -31). Each subunit consists of three structural domains related to each other by a pseudo 3-fold symmetry. Pairs of subunits form dimers, and each dimer is a tight complex, with a wide interface area between subunits, that accommodate two active sites (29 -31). The active site of the enzyme is made by both subunits. However, no information about the folding mechanism of MAT is still available. We have characterized the folding-assembly pathway of MAT III, the specific liver isoform, by identifying the intermediates involved in the process. Our kinetic and equilibrium experiments indicate that the folding of MAT III involves the formation of two folding intermediates, which were identified by their intrinsic fluorescence, ANS binding capacity, and hydrodynamic volume.

EXPERIMENTAL PROCEDURES
Materials-Columns and chromatography media were from Amersham Biosciences. Urea was from Invitrogen. All other reagents were from Sigma.
Site-directed Mutagenesis-A 1.2-kilobase fragment containing the rat MAT1A coding region (32) was subcloned into a pET vector. The resulting plasmid includes a 5Ј sequence that encodes for 6 histidine residues and a thrombin cleavage site in-frame with the rat liver MAT1A coding region. Mutants were obtained by inverse polymerase chain reaction according to the procedure of Pérez-Mato et al. (33). Mutants were identified by sequencing the complete MAT cDNA.
Expression and Purification of His-tagged MAT I/III R265H Mutant-MAT I/III R265H mutant was overexpressed in E. coli BL21(DE3) as previously described (33). Recombinant MAT I/III R265H mutant protein was purified from the bacterial cytosolic extracts by affinity chromatography on a Ni 2ϩ -Sepharose column equilibrated in 50 mM Tris/HCl, pH 8, 0.5 M NaCl, 75 mM imidazole following the procedures described by Pérez-Mato et al. (33). No changes in the fluorescence, CD spectra, oligomeric state, and enzymatic activity were observed after digestion of the recombinant proteins with thrombin (data not shown). Therefore all the assays were carried out using the Histagged recombinant proteins.
Purification of MAT from Rat Liver-MAT III was purified from rat liver according to the procedure described previously (34). Protein purity was more than 95% as estimated by SDS-PAGE (35).
Size Exclusion Chromatography-Protein samples (4.6 M) were analyzed using a Superdex 200 HR 10/30 column equilibrated with 50 mM Tris/HCl, pH 7.5, 150 mM KCl, 10 mM MgCl 2 in an AKTA FPLC (Amersham Biosciences). After sample injection (100 l), proteins were isocratically eluted at a flow rate of 0.8 ml/min. Protein elution was monitored by the absorbance at 280 nm. The elution volumes (ml) of the standard proteins were: thyroglobulin (669 kDa), 8 19. According to the elution volume of the standard proteins, the estimated molecular masses of native rat liver MAT III and R265H MAT I/III mutant were 90 and 41 kDa, respectively.
Equilibrium Denaturation Experiments-The buffer composition in all equilibrium experiments was 50 mM Tris/HCl, pH 8, 4 mM MgCl 2 , 150 mM KCl, and 3.6 mM dithiothreitol with different urea concentrations. All experiments were performed at 25°C with the protein concentration indicated in the figure legends and incubated in the presence of urea for at least 2 h. Unfolding was monitored by recording the fluorescence emission spectra at each urea concentration from 300 to 450 nm emission wavelength using an excitation wavelength of 280 nm in a PerkinElmer LS50B luminescence spectrometer. An emission wavelength of 321 nm is used in all plots, because this is the wavelength where differences were maximal. After recording the emission spectra, ANS was added from a 1 mM stock solution to make a final concentration of 10 M. When required, tripolyphosphate/magnesium from a stock of 50 mM was added to make a final concentration of 500 M. The fluorescence at 470 nm emission wavelength was recorded in both cases using an excitation wavelength of 350 nm. When unfolding was monitored by measuring the tripolyphosphatase activity, MAT III was incubated for 2 h at 25°C in 990 l of the above buffer. After this time, the reaction was started by addition of 10 l of tripolyphosphate/magnesium to make the final concentration 2 mM. At different incubation times, aliquots of 100 l were removed to determine the formation of inorganic phosphate according to the method of Lanzetta (36). For refolding experiments MAT III was unfolded in the same buffer with a final urea concentration of 8 M. After 2 h at 25°C refolding was started by diluting unfolded MAT III 20 times to obtain the desired urea concentrations. Refolding proceeded for 2 additional hours. Fluorescence values were normalized by the protein concentration and expressed as arbitrary units per nanomolar MAT III monomers. Enzyme activity was expressed as min Ϫ1 .
Refolding Kinetic Experiments-MAT III was partially unfolded for 2 h at 25°C in the same buffer indicated above with a final urea concentration ranging from 1.4 to 2.6 M. Refolding was started by a 1/10 dilution of the MAT III solution to obtain a final urea concentration of 0.26 M. Intrinsic fluorescence was monitored using excitation and emission wavelengths of 280 and 321 nm, respectively. When ANS binding was monitored all solutions contained 10 M ANS, and refolding was monitored using an excitation wavelength of 350 nm and an emission wavelength of 470 nm. The approximate dead time was about 10 s.
Limited Proteolysis and Mass Spectrometry Analysis-Limited proteolysis was performed by incubation of MAT III (0.5 mg/ml) with trypsin (0.01 mg/ml) in 10 mM ammonium bicarbonate, pH 8, at 25°C for different periods of time. Protein fragments were analyzed by SDS-PAGE. Mass spectrometry analysis was carried out with a Micromass MALDI-TOF spectrometer. After digestion, reaction mixture was 5-fold diluted with 0.1% trifluoroacetic acid and then spotted (1.5 l) onto the MALDI target, where it was mixed with an equal volume of matrix (cinnaminic acid). The analysis was performed using a laser energy of 70%, and M r was calculated using adenocorticotropic hormone and trypsin autolysis fragments as near point and internal standards, respectively.
Data Analysis-Equilibrium experiments were all fitted to a threestate model with a folding/association step and an additional folding step (Equation 1). As discussed below, the folding intermediate is assumed to be monomeric.
where N, I, and U are native, intermediate, and unfolded protein, respectively. K 1 and K 2 are the equilibrium constants for the folding/ association and folding steps, respectively. If the species present at each urea concentration are expressed as fraction (f), then the conservation of mass and the equilibrium constants can be expressed as follows, where P is the total concentration of MAT III subunits. From Equations 2 and 3, and solving for the fraction of each species we get the following equation.
The dependence of the equilibrium constants on the denaturant (D) concentration can be obtained form the free energy of unfolding ⌬G u where ⌬G u H2O is the free energy of unfolding in water, and m is the coefficient of dependence on the denaturant concentration.
The measured signals (S D ) depend on the species composition at each urea concentration according to the following equation, where S N , S I , and S U are the specific signal of native, intermediate, and unfolded protein, respectively.
The equilibrium experiments were fitted in the spreadsheet Microsoft Excel as follows. The equilibrium constants K 1 and K 2 were calculated for each denaturant concentration using initial estimate values of ⌬G u H2O and m. Then, the fraction of each species was calculated using Equation 4. The specific signal of each species was then calculated as the regression coefficients of the fraction of the species in a multiple linear regression analysis between the calculated fractions and the measured signals at all urea concentrations, using a built-in function of the program. This is like solving Equation 6 for all urea concentrations. The theoretical signal was then calculated using Equation 6. The agreement between calculated and measured signal was then evaluated by least square. Changing only ⌬G u H2O and m using an iterative procedure (quasi-Newton method) obtained the best fit. In this way, all the different types of signals measured could be fitted simultaneously to the same folding mechanism with only two variables per transition. When several conditions were evaluate at the same time, such as unfolding and refolding or different protein concentrations, average specific signals were used to calculate the theoretical signal. Sometimes the specific signal appeared to depend on the denaturant concentration. In such cases, a slope was introduced in the fitting process.
Kinetic data were fitted to single first or second order rate equations, where S(t) is the signal recorded at time t, Sϱ is the signal at t ϭ ϱ, A is the amplitude of the signal, and k is the rate constant.

RESULTS
Intrinsic Fluorescence-The equilibrium unfolding and refolding in the presence of urea was monitored by changes in the intrinsic fluorescence of MAT III at 321 nm (Fig. 1A). The unfolding of MAT III was a completely reversible process with two transitions. The initial drop in fluorescence was the result of a decrease in intensity, whereas the second one reflected a red shift of the spectrum without a significant change in fluo-rescence intensity (inset in Fig. 1). These data indicate the presence of a stable folding intermediate whose tryptophan residues are still buried within the protein. The midpoints for the transitions were 1.10 and 3.82 M urea using 0.28 M MAT III monomer (see thermodynamic parameters in Table I).
Equilibrium experiments performed with the mutant R265H, which is a monomer (37), also showed two transitions (Fig. 1A). The first transition took place at a lower urea concentration compare with wild type MAT III, whereas the second transition paralleled the one observed at 3.82 M urea. This result suggests that both proteins unfold via the same folding intermediate.
The second transition, corresponding to the unfolding of the intermediate, must be a unimolecular process, since the R265H mutant is a monomer. Thus, we studied the protein concentration dependence of the first transition. As shown in Fig. 1B, the transition was shifted toward higher urea concentrations with increasing MAT III concentrations. This result suggests that the transition observed at about 1 M urea represents the folding/association process between native MAT III and a stable intermediate.
Enzyme Activity and ANS Binding-Equilibrium experiments were also monitored by measuring tripolyphosphatase activity as well as ANS binding capacity. Both parameters showed a single transition at about 1 M urea ( Fig. 2A). ANS binds to large patches of hydrophobic surface exposed to solvent and has been widely used to probe folding intermediates. However, ANS seemed to bind the native conformation of the enzyme rather than the intermediate. We have determined an apparent binding stoichiometry of 1 mol of ANS per MAT III subunit (not shown), suggesting that ANS binds to a specific site. Recently, the crystal structure of ANS bound to a single site of a protein has been solved (38). A similar environment  Table I. was found in the MAT III structure at the entrance of substrates to the active site. Supporting the existence of such a binding site was the quenching of the fluorescence signal produced by the addition of the substrates of the enzyme ( Fig. 2A). Thus, enzyme activity and ANS binding are both probably probing the conformation of the same structural domain of the protein. This would explain the parallel loss of activity and ANS binding upon unfolding of the native protein.
Hydrodynamic Volume-The compactness of the species present at different urea concentrations was studied by gel filtration chromatography. In the absence of urea the elution volume of MAT III and R265H mutant corresponded to that of dimeric (14.3 ml) and monomeric (15.3 ml) forms, respectively (Fig. 3). However, as the urea concentration increased the elution volume of both proteins became closer until reaching the same intermediate value (14.4 ml) at about 1.25 M urea. At higher urea concentrations the decrease in elution volume was the same for MAT III and R265H mutant until reaching the value of unfolded monomer (13.1 ml). These data also support the presence of a common folding intermediate in the unfolding pathway of both proteins. This folding intermediate had a less compact structure than the monomer as indicated by its larger hydrodynamic volume.
Trypsin Accessibility-To gain information about the conformation of the folding intermediate, limited proteolysis experiments were performed. In the absence of urea MAT III was resistant to cleavage by trypsin for 1 h at 25°C (Fig. 4). However, in the presence of 1 M urea, a single proteolytic fragment was detected by SDS-PAGE. It was estimated to have about 3 kDa less than native MAT III. This is consistent with a single cleavage either at the N-or C-terminal end of the protein. The mass of the complementary fragment, as determined by mass spectrometry, was 3480.5 Da, which is consistent with a cleavage at the N-terminal side of lysine 33. This residue is located near the interacting interface between subunits. Thus, in order for this residue to become accessible to trypsin, a monomeric intermediate should be present.
Refolding Kinetics-To further characterize the transition between native MAT III and its folding intermediate, the refolding kinetics was monitored by changes in fluorescence intensity at 321 nm. To study the protein concentration dependence, different concentrations of MAT III (from 0.3 to 30 M) were incubated at 2.3 M urea. Under these conditions the folding intermediate represented over 95% of the protein. Refolding was started upon a 1/10 dilution. The same refolding kinetics was observed at all MAT III concentrations. Moreover, the data could be fitted to a single exponential but was poorly described by a second order rate equation, indicating that the traces shown in Fig. 5 corresponded to monomolecular reactions. Similar results were obtained when refolding was monitored by ANS binding (not shown). These results imply that in  (Table I). These results suggest that it is a folding intermediate that is not significantly populated in our equilibrium experiments. However, it must bind a large amount of ANS nonspecifically in order to be detected.

DISCUSSION
The overall unfolding pathway for a dimeric protein must begin with the folded dimer (N 2 ) and end with two unfolded monomers (2U). The way in which this process occurs depends on whether intermediates are present along the pathway. Numerous examples exist of compact monomeric or dimeric intermediates that are significantly populated during the equilibrium denaturation of a folded dimer (6 -9). However, far from being a general mechanism, the formation and role of these intermediates needs to be addressed in light of the protein under investigation. Here we describe the folding mechanism of MAT III, the liver isoform of a ubiquitous enzyme that synthesizes AdoMet. MAT III is a homodimer of a 43.7-kDa subunit (17,20) that organizes in three nonsequential ␣-␤ domains (29 -31). Although the MAT III structure has been recently resolved, little is known about its folding mechanism.
All our results from equilibrium experiments have been fitted to the following folding scheme,  (37) (and it is involved only in monomolecular reactions, not shown), the folding intermediate must also be a monomer. The first transition observed in the case of the mutant protein, however, took place at a lower urea concentration, indicating that the monomeric mutant is less stable than dimeric MAT III at the same protein concentration. Moreover, even in the absence of urea, the R265H mutant seemed to be in equilibrium with the folding intermediate, because no plateau was observed.
The stable folding intermediate may be a molten globule. The physicochemical properties of the folding intermediate are similar to those proposed for the molten globule (39). It has an intermediate compactness as indicated by its hydrodynamic volume, which is larger than the monomer, but much smaller than the unfolded protein. Its fluorescence spectrum indicated that tryptophan fluorescence was quenched (compare with native protein). However, tryptophan residues did not seem to be exposed to the solvent, because the emission maximum was the same as in the native protein. After its unfolding, a red shift of the spectrum was observed, indicating their exposure to the media. The folding intermediate was inactive and unable to bind ANS specifically (see below). There were no hydrophobic patches exposed to the solvent, because it did not bind ANS, indicating that the dimerization interface was not accessible. Lysine 33 is located more or less in the middle of the dimerization interface with its side chain buried within the protein. Thus, to be accessible to trypsin digestion the folding intermediate must be a monomer with the dimerization interface completely altered. This lysine residue appeared to be the most sensitive toward trypsin cleavage, even more than lysine residues at the surface of the protein, which suggests that the N-terminal region might be extended.
There is a specific binding site for ANS that may be located near the access to the active site. The apparent binding stoichiometry of 1 mol of ANS per MAT III subunit suggests that ANS binds to a specific binding site. Recently, the crystal structure of a binding site for ANS has been described (38). ANS is stacked in a hydrophobic cleft made by the sequence Pro-Gly at one side and the hydrophobic side chain of an arginine at the other side. A similar environment is also found in MAT III. The only Pro-Gly sequence in MAT III is located opposed to the hydrophobic side chains of Val-Gly-Ala. This last sequence is at the beginning of the loop that limits the access to the active site of the enzyme and that is probably involved in catalysis (40). The quenching of ANS fluorescence after addition of substrates supports the location of this putative binding site. Moreover, the loss of activity paralleled the quenching of ANS fluorescence, suggesting that they are probing the same structural domain of the protein.
There However, our data do not support all these requirements. Whereas the first requirement might be true for intrinsic fluorescence, it did not seem to hold for fluorescence due to ANS binding (Fig. 6). Then, accumulation of the kinetic intermediate would produce different kinetics (with losses of amplitude) when monitored by tryptophan or ANS fluorescence. Thus, the kinetic intermediate is most probably a monomer. There are several possibilities regarding the spectroscopic properties and the rate-limiting step. However, the simplest mechanism that explains our data with the minimum number of assumptions is a monomeric intermediate with spectroscopic properties (at least intrinsic fluorescence) like those of the native protein and whose formation (the folding step) is rate-limiting. Thus, no protein-dependent lag phases nor losses of amplitude, which were not seen, should be expected. It could be speculated that this kinetic intermediate would be the folded or very structured monomer. Thus, exposure of the large hydrophobic dimer interface would be energetically unfavorable so that association is expected to be very fast, maybe diffusion-controlled. This would also explain the high nonspecific binding of ANS of the intermediate.
We propose that both MAT III and the R265H mutant would share the two folding intermediates described here. However, once the kinetic folding intermediate would be formed the folding pathways of both proteins would diverge. In the case of MAT III, association of two monomers would take place rapidly. Since association is impaired in the case of the R265H mutant, an alternative, more compact monomeric conformation would be produced. This conformation would presumably occlude the mainly hydrophobic interface between subunits to stabilize a monomeric conformation.
In summary, our results indicate that MAT III unfolding is fully reversible with the following proposed folding mechanism, where U, I, and N denote unfolded, intermediate, and native protein, respectively, and the subscript k indicates that it is a kinetic intermediate. The subscript m and the asterisk indicate that it is a monomeric mutant with an altered native-like conformation. The stable monomeric folding intermediate has an altered non-native dimerization interface with no exposure of hydrophobic patches or tryptophan residues to the solvent. Although it is less compact than native MAT III, it is much more compact than unfolded protein. The kinetic intermediate present in the transition between the stable folding intermediate and native MAT III is most probably a monomer whose formation is rate-limiting and associates very rapidly to form native MAT III. The monomeric mutant R265H unfolds via the same stable folding and, probably, kinetic intermediates. Our data also indicate that ANS binds MAT III specifically at a site that is probably located near the access to the active site.